← Ascendy 한국어

meta

Using AI to write your résumé isn't cheating — so what do you read instead?

· Ascendy Engineering

TL;DR

About this piece. A first-person column drawn — via an interview (/interview) — from an operator who built a “don’t fabricate” portfolio tool while job-hunting. Throughout, I deliberately separate assertion (“don’t judge by the résumé alone — read the record”) from speculation (a vouching institution, etc.). Same record-as-trust vein as the story of a tool that builds a grounded portfolio from real GitHub work.

”Did you write this with AI?”

There’s a quiet suspicion going around hiring lately: “This cover letter — was it written by AI?” As if it’s something to be caught at.

But step back. Was the résumé a human wrote ever honest? We always exaggerated. “Led the initiative” was really “helped out next to the person who led it”; “improved by 30%” picked the measurement that flattered. A résumé is, by design, the document that packages you at your best. Whether AI polishes the sentences or not, a single résumé is a one-page claim anyone can inflate and forge.

So “was it AI?” misses the point. The cheating isn’t in the tool. The problem is the structure of judging a person by one unverifiable document.

So — read the record, not the résumé

The conclusion is simple. If you won’t trust the résumé, don’t judge by that one page alone. So what do you look at?

Look at the timestamped, accumulating record instead. Developers already have one — GitHub PRs and commits, what you built and when, stacked in time order. Yes, this can be gamed too. But touching up a one-page résumé once and forging years of consistent accumulated history are different difficulties. And whether that record was written by a human or an agent, the accumulated trail itself beats a single claim.

The point is to change the question from “who wrote it?” to “what actually accumulated?"

"That’s a developer thing”

The immediate objection: “GitHub? That’s for developers. I’m a marketer / PM / in sales — what accumulated record do I have?”

Fair — developers are furthest ahead right now. But the direction is the same: just as developers manage code, non-developers should keep a record of what they make. Campaigns you ran and their results, products and features you shipped, documents you wrote and published, deals you closed. Whatever it is, timestamping it as you make it becomes an asset.

“Couldn’t you fake that too?” You could. But it gives more chances for the lie to surface than a single résumé does. Someone who filled an accumulated record with lies tends to show it in the interview, and if not there, in a short trial of working together. Of course neither interviews nor trials are perfect detectors — they have false positives and negatives, and a timestamp alone doesn’t prove who did the work or in what context. The claim isn’t a guarantee; it’s that the surface you can verify gets wider.

The hiring criteria are changing

Until now we hired by proxy. Which school, which certificate, which company. We couldn’t see the real ability directly, so we read a signal that stood in for it.

That’s changing. What matters now is what this person actually built, and the data they’ve accumulated. Honestly, that was always the more accurate signal — the problem was the receiving side couldn’t process it. Reviewing years of activity for hundreds of applicants, by hand, was impossible and inefficient. So we leaned on the summary called a credential.

That constraint is starting to loosen. It’s now possible to process large volumes of data with agents in a short time — the cost of reading, cross-checking, and organizing hundreds of applicants’ accumulated records can drop sharply, even if it isn’t a fully proven workflow yet. And with it, the reason to lean on the summary shrinks.

It’s already happening

This isn’t far off. Look at the Japanese AI company Sakana AI (founded by David Ha, ex–Google Brain, and Llion Jones, a co-author of the Transformer paper). To be honest, they don’t skip the résumé entirely — you apply with a CV and cover letter. But the published applicant guide they put out (for research roles, co-authored by Llion Jones) states one core principle — “understanding over implementation.” Plenty of people can build something; far fewer can explain the reasons, limits, and improvements of their own design.

From here it’s my view: even if the résumé isn’t abandoned, the center of gravity is already shifting toward “what you actually built, and how well you understand it.” I think that’s a better fit for hiring capable people in the AI era. Not the verbal polish of a one-hour interview, but the output, the record of how it came to be, and the ability to explain why you did it that way — that is the person.

When the record itself becomes a resource

One last thing. Some people kept a record long before the AI era — of meetings, decisions, what they made, what failed. They weren’t recording in anticipation of this moment. Yet now, they hold the biggest advantage.

Because that record can be transplanted straight into a personalized agent. What you accumulated becomes your knowledge graph, working on your behalf and proving who you are. In the agent era, a record is a far more powerful resource than it ever was.

The open question (assertion vs speculation)

That’s my assertion — don’t trust the résumé, look at the record.

What follows is still speculation. The biggest obstacle is that much of the record is internal company data you can’t release. One possible path: a trusted institution that masks the data rather than exposing it, and vouches for it as a record both sides can trust — a kind of mechanical reference check. This is an idea, not an institution that exists.

And this direction has real costs — let me put them plainly. A long take-home assignment offloads unpaid labor; for someone already employed or caring for a child or family, that time is itself a privilege, so it can quietly favor whoever has spare time. Evaluation centered on the publishable record undervalues the work that holds things up from the inside — people who help others, coordinate, and do work that only matters internally have fewer externally visible artifacts like GitHub. And the moment “what’s recorded becomes the evaluation,” people start recording only what scores (Goodhart’s law) — performative activity rises, along with the pressure to constantly log and surveil yourself. Privacy, roles with no record, and qualities that don’t show in numbers (leadership, attitude) all remain. These aren’t solved — they’re the next problems to work on.

But the direction is clear. From an era that asks “did you use AI?” to one that asks “what have you accumulated?” A record etched by time is more honest than a one-page claim anyone can fabricate.

Authorship & citation: Written by Ascendy Engineering; quotable with attribution. Found something wrong? Let us know via a GitHub issue.

Tags: ai, hiring, career, opinion, future-of-work